On Thu, Jan 19, 2023 at 09:00:45AM +0100, Nuno Sá wrote:
> On Wed, 2023-01-18 at 08:37 -0800, Lars-Peter Clausen wrote:
> > On 1/18/23 07:49, Andy Shevchenko wrote:
> > > On Wed, Jan 18, 2023 at 07:22:30AM -0800, Lars-Peter Clausen wrote:
> > > > On 1/17/23 23:48, Andy Shevchenko wrote:
> > > > > None of the current users is using gaps in the list of the
> > > > > items.
> > > > > No need to have a specific function for that, just replace it
> > > > > by
> > > > > library available __sysfs_match_string().
> > > > Hm, I specifically remember adding this for a driver where there
> > > > were gaps.
> > > > One of the DACs. But it might be that the driver itself never
> > > > made it
> > > > upstream.
> > > I have checked all modules that have struct iio_enum and/or ("or"
> > > probably may
> > > not happen) IIO_ENUM() in them.
> > >
> > > It might be that I missed something.
> > I checked too, I can't find it either. The driver probably never made
> > it
> > upstream.
>
> Yeah, I also did a quick check and I could find it in one adc (most
> likely we have more downstream users of this) that did not make it
> upstream. Eventually, we want to have it upstream but the ABI using the
> gaps can arguably be dropped...
>
> Anyways, from my side I'm fine with this change. We can revert it if we
> ever have a real user for this. I'll just have to be careful when
> updating ADI tree (but that is our problem :)).
We usually do not keep a dead code in the kernel, and handling gaps is
a dead code. And yes, we always can return to that when we have a user,
most likely as a part of the generic library and not just IIO.
--
With Best Regards,
Andy Shevchenko
